PERFORCE change 135023 for review
Warner Losh
imp at FreeBSD.org
Fri Feb 8 00:29:26 PST 2008
http://perforce.freebsd.org/chv.cgi?CH=135023
Change 135023 by imp at imp_mbp on 2008/02/08 08:28:24
More mind weasils ripping apart the flesh of my brain.
Affected files ...
.. //depot/projects/mips2-jnpr/src/sys/TODO#16 edit
Differences ...
==== //depot/projects/mips2-jnpr/src/sys/TODO#16 (text+ko) ====
@@ -6,19 +6,24 @@
o Can we remove pre-ansi stuff
o unify cpu.h vs cpureg.h
o Look for old vestiges of the array of args for p->p_regs[] and also comments
- about same. Struct reg, fpgreg seem OBE too.
+ about same. Struct reg, fpgreg seem OBE too. On the other hand, keeping
+ constants unified between the two different conventions is a pain and we are
+ having some data corruption issues for reasons unknown on the MALTA
+ simulation at the moment.
o Look at the differences between the trampoline code between the two versions
as this may lead to problems. Specifically, we have sigcode encoded
into locore.S, and some code in sigreturn in pm_machdep.c. Not sure where
things are in libc.
o Don't forget John Baldwin's comments on the user switch stuff. XXX add more
- here XXX
+ here XXX This is mostly down to AST things.
o What's defined in assym.s and param.h varies somewhat between the two
code strains. Figure out why and how to resolve.
o machdep getmemsize() really is OBE. Except for the mips_wired_tlb
variable setting, which is related to something that was omitted from
the juniper-mips dump.
-o cpu init
+o cpu init ??? DONE ????
+o we have a lot of memory from 0x80001000-0x81000000 that's currently
+ wasted. We should feed it to the vm system so as not to waste it.
o Double check last merge from mips2 -> cavium code. Something seems wrong
in one of the edits. Gonzo suggests looking at DO_AST macro.
o Look at what the macro to set the per cpu data pointer will be like, and also
More information about the p4-projects
mailing list